There once was the T800 transputer. Much renowned for the four very fast communication links, allowing networks of transputers.

Where are they now?

Multicore PC processors got introduced because the manufacturers don't know how to step up the clock speed any more. The advantage is that multiple applications can run concurrently without affecting each other - when IE or flash hangs and hogs 100% of a core, the rest will continue to work. But the disadvantage is that the software development costs for getting a single application to take advantage of multiple cores is high.

Multicore is very much a hack to get around a quite hard performance bottleneck.

Any developer who are able to produce a tightly coupled multiprocessor system (i.e. where the processors don't just stream data one way but regularly needs to interchange information in both directions) should also be able to handle interprocess communication within a single processor core. Good peripherial hardware devices can greatly help with I/O latencies, letting the majority of the time being dedicated to making decisions based on existing stimuli.

Another thing is that a processor that halts draws way less energy than a processor running instructions. So it is possible to produce 1GHz (or faster) microcontrollers without blowing the energy requirements. The peak performance is just used for bursty sequences, while the processor spends the majority of the time sleeping.

In the end, RTOS-driven or highly interrupt-driven software is quite easy to write and also quite inexpensive compared to general tightly coupled multiprocessor systems. It's only when you can get a way with a master and multiple slaves that you can design inexpensive multiprocessor or multicore solutions without the development time and maintainance costs exploding.